SergioLuis
commented
6 years ago First, my responses to some of the comments:
- About 927dd0f (Change the PC to Int to avoid conversions)
IMemory
- About a4b006b (Add a 'PaddedMemory' to emulate original memory)
The PaddedMemory class is only to avoid other less elegant solutions such as changing the PC on each call. The specialized classes are there to take advantage of ByteArray and ShortArray, which are converted to byte[] and short[] instad to a generic Array<Byte> or Array<Short>. If we didn't use specialized classes, we could get rid of ByteMemory and ArrayMemory and just use a generic Memory<T> : IMemory<T>. Because we rely on the interface instead of on the implementation, we can later change this effortlessly.
- About 77f1b5d (Refactor memory classes to receive an array)
Fixed the EOF of IMemory.kt and README.md at bc87cce
- About 9f994ab (Define the IRenderer interface) and 375285f (Add method to the IRenderer to clear the display)
A renderer is not a window/screen/surface. Is a renderer.
Correct.
Clear should be called ever before calling any rendering functions.
Incorrect. There is an specific instruction to clear the display, is none of the emulator bussiness.
This allows implementing things like double buffering, vsync and such.
That is none of the emulator bussiness. The client can implement a double-buffer just with the render method.
Maybe we don't need a renderer interface in first place
What do you suggest to issue draw commands in a platform-agnostic way?
- About 758ee55 (Minor refactor to IClockDivider and subtypes)
Converted the function getMsLeft to the read-only property msLeft at 81d0c04.
Renamed the function init to reset at eb56498
About this:
...should be stated if calling init/reset before using the object is recommended or required.
I think it should be REQUIRED before entering the trigger-sleep while-true loop. BUT, as it is an internal part, I don't want to add a way to enforce it.
- About e51b0aa (Add an interface to sleep the thread)
I would prefer to rename every ms appearance to milliseconds (for consistency), but I complied with the review comment at 7c3cd2b
- About 7f0539f (Remove unused method on IClock interface)
Applied review comments at 76431e2
- About 66b7649 (Simplify instruction reading)
Simplified instruction reading at 9b9c4d8
- About 9815f38 (Implement and test the ldVxByte instruction)
Renamed the byte val to b at a1b6015
Also:
- 927dd0f Delete MyClass.kt example file
The general execution infrastructure is in place. The code is pretty self-explanatory, but requires the reader to be familiar with how the original system worked. You can check out the online available documentation.
Right now, the emulator is able to load a ROM and start the execution (single-threaded). The
JP addr,CLSandLD Vx, Byteinstructions are already implemented and tested as an example on how to do so.@jhvargas3112 @ugedo please get familiar with the code and ask any doubts you have. I'll divide the instruction set soon and everybody will get to work on this core.
I created the Pull Request in order to get a formal review on the code. Some work is still to be done (for example, loading an snapshot), but I think we should leave that for later, as this task has gotten out of hand in invested hours terms.